1. Field of the Invention
The present invention relates to multi-tasking micro-processing systems and, more particularly, to multi-tasking systems employing networked stand alone and independently operating micro-processing units each of which perform one or more tasks for manipulating electronic data.
2. Related Art
The processing of large amounts of electronic data is becoming more and more common among business concerns. Efficient manipulation of the electronic data is critical because, even with ultra fast computers, the magnitude of electronic data to be manipulated places limits on throughput.
With reference to FIG. 1, a prior art multi-tasking computing system 1' employs a central managing unit 10' which receives or monitors incoming electronic data. The system 1' also includes a plurality of processing units 12' coupled to the central managing unit 10'. A memory 14' may also be employed to store data as needed.
The multi-tasking system 1' of FIG. 1 is sometimes called a Massively Parallel Processor (MPP) system and is used, for example, in the IBM SP2, NCR, and Cray systems. An MPP system 1' controls the processing units 12' at the basic operating system level. Thus, the central managing unit 10' controls which processing unit 12' performs which jobs in manipulating the electronic data. In this way, the throughput of the entire system is improved because each processing unit 12' performs only a part of the overall work required.
Unfortunately, the multi-tasking system 1' of the prior art has several disadvantages. For example, the central manager unit 10' imposes overhead burdens on the system in terms of time, cost and complexity while performing no value added work on the electronic data. Further, the processing units 12' of an MPP system 1' typically perform relatively small units of work as compared to the total work required.
Another disadvantage of the prior art system is that the hardware must be custom designed and usually requires special network switching technology. Consequently, scalability of the system may become problematic because the central managing unit 10' may require custom hardware and/or software configuration changes when other processing units 12' are added to the system 1'.
Accordingly, there is a need in the art for a multi-tasking computing system which utilizes a standard network over which a plurality of processing units communicate, which system is readily scalable.